At Indian Institute of Management (IIM) Ranchi, I supported the end-to-end data science workflow and benchmarked FMLP-Rec against five sequential models across four large-scale transactional datasets.
I stress-tested model robustness with synthetic noise and evaluated scalability across latency, throughput, and sequence-length sensitivity. I also documented reproducible evaluation protocols.
In my EV Charging Demand Analytics & Overload Prediction project, I built a pipeline using more than 259,000 sensor records and trained a Random Forest classifier with a ROC-AUC of 0.94. I also grouped stations with KMeans and shared insights through a Power BI dashboard.
I’ve built machine learning projects for delivery-delay prediction, marketing response, and credit-risk scoring. My work has included deploying a CatBoost model as a FastAPI service with Docker and using model explainability to make credit-risk outputs interpretable.
